[WSJ] TSMC is shipping 20nm SOCs to Apple

[AtenRa]

Yields for Apple was 50% back in late April. It have increased a good amount by now, but there is no way they are near 100%. 60-70s maybe?

Got the link for that ??

 

[Homeles]

What matters is what's available in the shops and when.

This is silly. You're trying to compare TSMC's and Intel's process, which went into production at roughly the same time. Ship times are similar as well. After it's shipped, it's got nothing to do with the process.

 

[ShintaiDK]

The first SOC customer after Apple will be Qualcomm. And they will first start their production at TSMC in 3-6months. MediaTek will do it in 6-9 months from now. So I doubt yield is anything to write home about. And price is skyhigh.

Also remember binning. The chips may not be defect, but that doesnt mean they are useable.

 

[jpiniero]

iPhone 6's release is sometime in September. So yeah about 2 months from today. Apple's initial order for the iPhone 6 is 68 million, so they better get going.

If this is the first time getting 20nm silicon back from TSMC, I would be very surprised to see it in a product 90 days from now. Apple and TSMC would have to pull off a miracle.

It would have to be finished chips obviously, not first silicon.

 

[Fjodor2001]

Well both may come out at the same time give or take one or two months.

Yes, so TSMC might have chips at a later node than Intel in the shops for 1-3 months. That's all.

 

[Phynaz]

WOAH... what is this black magic!

Now that you know the secret handshake we have to kill you.

 

[ShintaiDK]

So for a few months, TSMC will be on a later node than Intel. 🙂

Yes, so TSMC might have chips at a later node than Intel in the shops for 1-3 months. That's all.

Losing confidence so fast?

May have yes, since it could just as well be after Intel. And only because a company like Apple is willing to pay a price for a SOC that is most likely higher than a GK110 chip.

 

[mavere]

Got the link for that ??

From early March:

"
Stock implications
TSMC
20nm yields for Apple SOC are already hitting 50%, much better than where 28nm was at the similar time of ramp. Wafer out should hit 6-7k in May and reach 25-27k by July/August. SoC yield is typically more difficult to stabilize than Graphics or FPGA so this is indeed a strong result.
"

 

[Cloudfire777]

Oh yeah, it was in March. Sorry about that

 

[Exophase]

Considering your statement. you must know when both products hits the shops. Do tell.

It's so strange, people keep acting like TSMC 20nm products are not actually out in things already.

 

[Roland00Address]

Yields for Apple was 50% back in late April. It have increased a good amount by now, but there is no way they are near 100%. 60-70s maybe?

I know there is no way the yields are close to a 100%, I just picked that number to show that apple needs a lot of volume from TMSC over a 3 month period, for most phones are bought right when the phone is released. Can TMSC handle all this volume at once? Is there 20nm that stable and they have enough capacity? I ask for we only recently started seeing 20nm tmsc products hit the market, as in the last month.

 

[DeathReborn]

Could these 20nm chips (if they exist) be for the much talked about but never seen iWatch?

 

[Homeles]

It's so strange, people keep acting like TSMC 20nm products are not actually out in things already.

Right now, it's just Qualcomm's modem. While technically in competition with Intel... it's not like Intel's now losing out much more than they already have been in that market. Apple's SoCs will likely be the next product we see, which aren't in direct competition with Intel.

So it's just not really worth mentioning, at least at the moment. Some customers paid extra and put up with lower yields to be first... great, so what? Intel's not any worse off. And 20nm is awfully boring, and of questionable usefulness to most fabless companies. TSMC has some sizable density lead over Intel... in a single analog component that doesn't scale well anyway. That density doesn't mean anything if your wafer costs skyrocket to the point of regressing cost/transistor, as has been frequently reported, or at least not making much progress. It's almost a Pyrrhic victory.

The upcoming FinFET nodes are far more important. Also far more important are the architectural-level innovations that Intel's competitors will be making, like Nvidia's Maxwell. 20nm itself hasn't changed anything, nor will it.

Intel's got 99 problems, but TSMC's 20nm ain't one of them.

 

[jpiniero]

Could these 20nm chips (if they exist) be for the much talked about but never seen iWatch?

It's definitely for the iPhone 6.

 

[NostaSeronx]

I think people will be surprised how much stuff is on 20-nm LPM at GlobalFoundries and Samsung.

 

[Arachnotronic]

Bah, why isn't anybody excited at what the A8 will bring?

The A7 was really neat, so personally, I can't wait to see the A8 in all of its 20-nanometer glory. Let's see what Apple/TSMC can do. I'm expecting that both companies will execute well.

 

[Exophase]

Bah, why isn't anybody excited at what the A8 will bring?

I don't want an iPhone 😉 I admire the quality of their CPU designs but it's kind of frustrating that they get to keep it to themselves - it's the first broadly used CPU family in a long time that's been locked to one consumer platform (especially if you don't count consoles)

 

[Arachnotronic]

I don't want an iPhone 😉 I admire the quality of their CPU designs but it's kind of frustrating that they get to keep it to themselves - it's the first broadly used CPU family in a long time that's been locked to one consumer platform (especially if you don't count consoles)

You will be assimilated? Resistance is futile? 😉

 

[jdubs03]

The A8 should be a solid follow up, maybe on par with the transition to the A7, my guess is the A9 could follow the same pattern. Though I don't expect it to get a large bump in those cryptography scores ala A7. But it should be a solid bump, 20% perf increase just from 20nm, 40-50% improvement overall I'm thinking. That's 2100 single-thread score for geekbench 3 (for 50%).

the benchmark race continues...

 

[blackened23]

The A8 should be a solid follow up, maybe on par with the transition to the A7, my guess is the A9 could follow the same pattern. Though I don't expect it to get a large bump in those cryptography scores ala A7. But it should be a solid bump, 20% perf increase just from 20nm, 40-50% improvement overall I'm thinking. That's 2100 single-thread score for geekbench 3 (for 50%).

the benchmark race continues...

It's hard to quantify Apple in the benchmark race since they're in their own ecosystem, iOS. They're really not competing on a SOC level with qualcomm at all. They're competing on a device level with Samsung, Nokia, etc, etc. Apple is generally very capable of creating hardware that creates a great experience (and yes, I do think their products are very good). But, again, in terms of the "benchmark race" you can't really quantify Apple in any metric except battery life? I understand the A7 has dominated benchmarks for some time, but it's hard just hard to compare cross platform numbers. They don't really equate to each other in apples to apples comparison at all times. I can see comparing android SoCs, but Apple competes (again) on a device level, no a SOC level. It isn't like Apple sells their SOCs to other companies for android use. Whereas qualcomm competes on the SOC level and not the end device level.

 

[jdubs03]

It's hard to quantify Apple in the benchmark race since they're in their own ecosystem, iOS. They're really not competing on a SOC level with qualcomm at all. They're competing on a device level with Samsung, Nokia, etc, etc. Apple is generally very capable of creating hardware that creates a great experience (and yes, I do think their products are very good). But, again, in terms of the "benchmark race" you can't really quantify Apple in any metric except battery life? I understand the A7 has dominated benchmarks for some time, but it's hard just hard to compare cross platform numbers. They don't really equate to each other in apples to apples comparison at all times. I can see comparing android SoCs, but Apple competes (again) on a device level, no a SOC level. It isn't like Apple sells their SOCs to other companies for android use. Whereas qualcomm competes on the SOC level and not the end device level.

For sure, and I only used one benchmark at that. I was just trying to paint a picture of what we could expect from the next iPhone/iPad. It is unfortunate that there is a disconnect due to Apple's ecosystem customized only for iOS. But say if the score were 2100, it would give us at least a comparison point against the A7, then all the other benchmarks will fill in the picture, we also get to see what PowerVR can do with their Rouge GPU solution (GX6650?)

 

[jhu]

I don't want an iPhone 😉 I admire the quality of their CPU designs but it's kind of frustrating that they get to keep it to themselves - it's the first broadly used CPU family in a long time that's been locked to one consumer platform (especially if you don't count consoles)

It's only been 8 years since the Apple used the PowerPC 970. I'm just disappointed that no one's been able to port Linux to any A7-based devices. Probably has something to do with having a locked bootloader. But then, Linux support isn't all that great on the devices where the bootloader has been cracked.

 

[witeken]

Bah, why isn't anybody excited at what the A8 will bring?

Because it will be substantially worse than competing SoCs (Core M).

 

[NostaSeronx]

Are you guys sure what is coming out is not the A7x but the A8?

Why can't it be the A7x with;
Quad-core Cyclone?
GX6650?

 

[AtenRa]

From early March:

"
Stock implications
TSMC
20nm yields for Apple SOC are already hitting 50%, much better than where 28nm was at the similar time of ramp. Wafer out should hit 6-7k in May and reach 25-27k by July/August. SoC yield is typically more difficult to stabilize than Graphics or FPGA so this is indeed a strong result.
"

nice thanks